FWD » Core Development » User Interface

I did run this with updated text metrics, so the sizing problem shouldn't be due to that.

The differences resolved in 1811s revision 11011.

The checkin also adds proper handling of window caption close button.

Note that a regression broke all the keyboard input, this is being solved.

1811s revision 11014 fixes an unhandled exception when alert-box closed through the window caption button.

When you say devsrv01, did you mean winsrv01?

Greg Shah wrote:

When you say devsrv01, did you mean winsrv01?

Yes :-).

Hynek Cihlar wrote:

Greg Shah wrote:

When you say devsrv01, did you mean winsrv01?

Yes :-).

Actually I meant windev01. When you connect remotely do you maybe rescale the resolution?

Sorry, my mistake too, I also meant windev01. :)

I did not intentionally scale the resolution, but it may have happened anyway. If the size is right, then we won't worry about that issue.

Did you look at the changes in #3048? How do those affect this task?

Greg Shah wrote:

Sorry, my mistake too, I also meant windev01. :)

I did not intentionally scale the resolution, but it may have happened anyway. If the size is right, then we won't worry about that issue.

Did you look at the changes in #3048? How do those affect this task?

I don't think the changes in #3048 affect this task.

Hynek Cihlar wrote:

Eugenie, when I rasterize alert-box icon using SharpPNGTranscoder I get a very distorted output. But if I use the unmodified Batik PNGTranscoder directly, I get the expected output. See below.

What do you advise?

The SharpPNGTranscoder just turns anti-alias off. It was introduced for drawing all solid icons like scrollbar buttons. And this is how the batik package works with image.

What is the original image? What is your goal? You need to have the image without anti-alias?

If you need antialias to be off and want to have some predefined picture in P2J the approach is:
- take original picture
- make a change to original picture to adjust the final result.

There ane no shortcuts here. You will have to find out how the changes in original picture produces the final result. Pixel by pixel.

The better way if the image size is constant I guess is to use simple bitmap. And no BATIK or other third party image handling tools.

2770a rebased against trunk 11012, the branch is now at rev 11014.

Can you provide an update on this work?

Greg Shah wrote:

Can you provide an update on this work?

The core changes are in place. And the issues above addressed.

TODOs are, retest after rebase against trunk 11012 and minor code cleanup.

2770a rebased against trunk 11015, the branch is now at rev 11017.

Hynek Cihlar wrote:

Greg Shah wrote:

Can you provide an update on this work?

The core changes are in place. And the issues above addressed.

TODOs are, retest after rebase against trunk 11012 and minor code cleanup.

There is one more regression, the window can be resized only by its right/bottom edges, but not the top/left ones. I am working on this right now.

2770a rebased against trunk 11016, the branch is now at rev 11024.

Is this ready for code review?

Please review 2770a revision 11025.

The revision resolves the ALERT-BOX layout issues mentioned in the above notes. It also resolves Swing window resizing issues - resizing cursor sometime not appearing, sometime disappearing unexpectedly, window resizing sometime moving the whole window. Plus other related fixes.

The revision passes GUI regression tests. Since the changes are only in GUI code, no ChUI regression testing is needed.

Greg Shah wrote:

Is this ready for code review?

Revision 11025 is.

Eugenie Lyzenko wrote:

The better way if the image size is constant I guess is to use simple bitmap. And no BATIK or other third party image handling tools.

Eugenie, please review 2770a revision 11025. It contains changes related to the subject above.

2770a 11025 also resolves #3067.

Code Review Task Branch 2770a Revision 11025

I'm generally fine with the changes.

1. In WindowBorder, please put a line break before the implements.

2. How does the AbstractContainer change affect ChUI?

Eugenie, please review 2770a revision 11025. It contains changes related to the subject above.

Three considerations:
1. Did you classify all cases when the aliasing must be on and when must be off? We need to avoid 4GL - P2J visual mismatch.
2. In AlertBoxGuiImpl the icon image coordinate has hardcoded coords dependent on PADDING character unit related coord. Why not to just center the image withing rectangle? What if the font size is changing, will the image position be OK.
3. Your drawImage() does not use image transparency, is it intentional? Does the 4GL show the same behavior?

Greg Shah wrote:

Code Review Task Branch 2770a Revision 11025

I'm generally fine with the changes.

1. In WindowBorder, please put a line break before the implements.

Resolved in 11026. I also added one more javadoc.

2. How does the AbstractContainer change affect ChUI?

The changed method is only executed in GUI context.

Eugenie Lyzenko wrote:

Eugenie, please review 2770a revision 11025. It contains changes related to the subject above.

Three considerations:
1. Did you classify all cases when the aliasing must be on and when must be off? We need to avoid 4GL - P2J visual mismatch.

There is only one case ATM where antialiasing needs to be ON, and that is the ALERT-BOX icons.

2. In AlertBoxGuiImpl the icon image coordinate has hardcoded coords dependent on PADDING character unit related coord. Why not to just center the image withing rectangle? What if the font size is changing, will the image position be OK.

The icon position depends on Windows message box font size theme settings.

3. Your drawImage() does not use image transparency, is it intentional? Does the 4GL show the same behavior?

Although the icons use transparency, I didn't have a need to turn that flag on. The source SVGs have transparent backgrounds (they don't contain any drawings) and so Batik rasterizer creates the rastered images with correct transparency.

For which cases is the transparency flag needed anyway?

Please merge 2770a to trunk.

2770a merged to trunk as revision 11017. The branch has been archived.

Please review task branch 2770b revision 11018, it fixes GUI regressions caused by assigning widget IDs to all BorderedPanelGuiImpl instances.

Code Review Task Branch 2770b Revision 11018

The changes are fine.

Please add the embedded mode fix to this branch (when you have it ready).

2770b revision 11019 fixes main window border being drawn even when the driver doesn't support window decorations. Please review.

Code Review Task Branch 2770b Revision 11019

The changes are fine.

2770a rebased against trunk 11019, the branch is now at rev 11022.

2770b revision 11025 resolves regressions introduced by 2770a, fixes selection of next active window when modal window with no visible owner is being hidden and attempts to fix Swing JFrame sometime not focusing when JFrame.requestFocus() called right after the window is made visible.

GUI regression tests passed, ChUI regression testing is not needed as the changes only affect GUI code.

Please review.

Code Review Task Branch 2770b Revision 11025

I'm fine with the changes.

Please merge to trunk.

2770b merged to trunk as revision 11020. The branch was archived.